Re: Bug#231235: arabtex font installation is broken

[Florent Rougon <f.rougon@free.fr>]

Hi,

>> I guess you mean mktexupd.
>> 
> Both scripts seem to be written for the same purpose. texhash is
> called at the end of updmap. This is documented in the manpage.

mktexupd is used to add a single file to a ls-R file. It avoids the need
to actually perform an ls -R when this is not necessary.

But it is true that texhash is called at the end of updmap unless
disabled by specific options. I had not found it since I was looking for
the mktexlsr string...

> If this is really true one should call updmap with the option
> --nohash to prevent it from calling texhash twice.

I don't think so, because updmap could create map files (as opposed to
only updating them) if used on a new installation, or after one of the
generated map files has been deleted, or if a new generated map file is
introduced in a new updmap version.

To sum it up:
  - if you don't run mktexlsr after update-updmap and before updmap,
    updmap may fail to find any of updmap.cfg (if it was missing in
    ls-R, unlikely but could happen) and the newly added map files (very
    likely to happen);
  - if texhash is not run at the end of updmap, you may end up with an
    ls-R that does not contain all of the generated map files such as
    psfonts_t1.map.

The first call could be avoided if the maintainer scripts called
mktexupd for every map file added by the package and if update-updmap
called it to make sure updmap.cfg is in ls-R. However:
  1. This requires maintainers to be careful;
  2. This will fail to "help" a bad TeX admin that forgot to run
     mktexlsr... yes, this one is a bit far-fetched but might help avoid
     bogus bug-reports...

So, if maintainers of TeX-related packages are careful and if one line
is added to update-updmap, it is possible to avoid an extra run of
mktexlsr. I am not sure it is worth the risk of buggy packages, but I
may be convinced.

-- 
Florent